Northern Tibetan Plateau cooling and aridification linked to Cenozoic global cooling: Evidence from n- alkane distributions of Paleogene sedimentary sequences in the Xining Basin

The Xining Basin on the northeastern Tibetan Plateau holds the longest continuous Cenozoic stratigraphic record in China. The sequence record contains considerable information on the history of Tibetan uplift and associated climatic change. In particular, high resolution n -alkane biomarker proxy and pollen records have been obtained from the Paleogene sediments of the Xiejia section of the basin. A combination of the n -alkane and palynological records reveals that the paleoclimate in the Xining Basin ex-perienced a long-term cooling trend from 50.2 to 28.2 Ma with a distinctive ecological event spanning 37.5 to 32.7 Ma. Since this ecological event, a vertical zonation of vegetation from lowland arid grasses, to middle-elevation subtropical broad-leaf plants, to high-elevation coniferous trees was established. We interpret that these changes in climate and vegetation were probably responses to a combination of long term global cooling since the Eocene climatic optimum and uplift of the surrounding mountains on the northern Tibetan Plateau in the early Cenozoic.

The Xining Basin on the northeastern Tibetan Plateau holds the longest continuous Cenozoic stratigraphic record in China. The sequence record contains considerable information on the history of Tibetan uplift and associated climatic change. In particular, high resolution n-alkane biomarker proxy and pollen records have been obtained from the Paleogene sediments of the Xiejia section of the basin. A combination of the n-alkane and palynological records reveals that the paleoclimate in the Xining Basin experienced a long-term cooling trend from 50.2 to 28.2 Ma with a distinctive ecological event spanning 37.5 to 32.7 Ma. Since this ecological event, a vertical zonation of vegetation from lowland arid grasses, to middle-elevation subtropical broad-leaf plants, to high-elevation coniferous trees was established. We interpret that these changes in climate and vegetation were probably responses to a combination of long term global cooling since the Eocene climatic optimum and uplift of the surrounding mountains on the northern Tibetan Plateau in the early Cenozoic. The Tibetan Plateau is the most extensive region of elevated topography on Earth. Its global significance does not lie simply in the initiation and development of Asian aridification and monsoons [1][2][3][4][5][6][7][8], but also in ocean chemistry and global climate [9][10][11][12][13]. Thus, studies of the evolving paleoenvironment and paleoaltitude history of the Tibetan Plateau since the India-Eurasia collision can benefit our understanding of the link between tectonics and long term climate change. Some previous studies have focused on the early elevation history of Tibet to provide constraints on the timing and *Corresponding author (email: fangxm@itpcas.ac.cn) magnitude of the early surface uplift of the region based on indicators of paleotopography in the geological record [14][15][16][17][18][19]. However, early Cenozoic elevation in the Tibetan Plateau is still poorly understood and documented. Moreover, paleoaltitude studies have focused on southern and central Tibet [8,14,16,17,20,21]. Such studies for northern Tibet are limited by poorly constrained ages for a long and continuous sequence of Cenozoic sediments. The lack of good age constraints on some deposits precludes the ability to determine the precise timing of surface uplift in northern Tibet. Further studies are clearly needed, particularly of northern Tibet, using more direct elevation proxies to chart the distribution and area of the uplifted plateau at a range of elevations throughout the Neogene.
Along the northeastern margin of the Tibetan Plateau, the Xining Basin holds the longest continuous Cenozoic stratigraphic record in China. The exceptionally long (52.0 to 17.0 Ma) continuous sedimentary succession in the basin provides an excellent opportunity to study the sedimentary signature of tectonic and climatic processes from the Eocene to the Oligocene. Stratigraphic, biostratigraphic, and magnetostratigraphic studies provide a basis for reconstructing the Cenozoic tectonic and climatic evolution of the Xining Basin with respect to adjacent regions of the northeastern Tibetan Plateau [22][23][24][25][26].
Reconstructing paleoenvironments using organic compounds constitutes a major field in the application of organic geochemistry. Plant n-alkanes from leaf waxes can be preserved in the geological record for a long time and register information on paleovegetation changes during historical periods. Analysis of those preserved in sediments is an established method for reconstructing the composition of past vegetation and thereby paleoenvironmental conditions [27,28]. The n-alkane ratio method of reconstructing paleovegetation and paleoclimate is advantageous because it is simple to perform and the size of samples required for n-alkane analysis is small. The strongly hydrophobic nature of the n-alkanes in sediment and their affinity for clays makes them immobile. Molecular stratigraphic approaches using plant n-alkanes have been employed widely for marine sediments [29][30][31], lacustrine and peat sediments [32][33][34][35][36] and loess-paleosol [28,[37][38][39]. However, plant n-alkanes preserved in the Xining Basin have not been used to deduce paleoenvironmental information. In this study, we present well-dated organic geochemical and pollen data from the well-studied Xiejia section of the Xining Basin. Note that the interpretation of paleovegetation patterns from pollen can be problematic because of allochthonous input and long distance transport of some particular plant pollens are not easily preserved. The combination of organic geochemistry and classical pollen analysis has proven to be especially effective in detailed regional reconstructions of paleoenvironmental and paleoclimatic history [40][41][42]. Thus, we have combined n-alkane proxies with pollen records to reconstruct the paleoclimate history of the basin, thereby providing new insight into the temporal relationship between the northern Tibet uplift and global climate.

Study area
The area of interest in this paper is located south of Xining on the northeastern part of the Tibetan Plateau in northwest China (Figure 1). The samples come from the Xiejia section (101°52'E, 36°31'N, 2250 m above sea level (a.s.l.)) in the Xining Basin. The basin is the transition zone from the Tibetan Plateau to the Loess Plateau, which is confined by Dabanshan to the north, Lajishan to the south, Riyueshan to the west, and a small uplifted block to the east. The elevations of the surrounding mountains are > 3000 m a.s.l. In contrast, the elevation of most of the area inside the basin ranges from 1750 to 2600 m a.s.l. The mean annual temperature (MAT) and precipitation are 5-6°C and 350-500 mm, respectively. Precipitation occurs mainly in summer. The native vegetation in the present basin has a dramatic vertical variation because of the large elevation differences between the valleys and surrounding mountains. Stipa bungeana, Stipa glareosa and Artemisia gmelinii are distributed mainly in areas below 2600 m. Picea crassifolia and Picea wilsonii are not abundant and are distributed mainly on shady slopes at 2600-3000 m. Shrubs, including Potentilla fruticosa, Salix oritrepha and Rhododendron, grow in areas above 3000 m. The alpine meadow plant, Kobresia bipartite is widely distributed in areas above 3200 m [43].
Eocene to mid-Miocene deposition in the Xining Basin was characterized by the slow accumulation of lacustrine saline playa to distal alluvial fan deposits [25]. The 819-m-thick Xiejia section can be divided into six formations, from oldest to youngest: Qijiachuan (thicknesses 20-69 m), Honggou (69-224 m), Mahalagou (224-504 m), Xiejia (504-667 m), Chetougou (667-778 m) and Xianshuihe (778-819 m). The lower part (0-450 m) of the section consists of roughly regular alternations of laterally-continuous gypsum layers and red mudstone intervals of lacustrine saline playa. The gypsum layers occasionally preserve greenish-grey lacustrine mud. The upper part (450-819 m) of the section is characterized by red beds of mudstone and siltstone from flood plains and distal alluvial fans, occasionally intercalated with thin layers of fluvial channel sandstone deposits [25]. A final gypsum layer occurs at about 500 m in the measured section ( Figure 2). The age model for the section has been established on the basis of magnetostratigraphy [25]. This study focuses on the portion from the top of the Honggou Formation to the bottom of the Xiejia Formation, from ~70.5 to 546 m in the section, corresponding to a time interval of ca. 50.2 to 28.2 Ma ( Figure 2).

Samples and methods
We collected 43 samples for geochemical analysis, with an interval of ~10 m from the Xiejia section (see stratigraphic positions, ages and sequences in Figure 2; Table 1). To better constrain the existing record in time and provide a new detailed geochemical analysis, sampling for plant n-alkanes was performed at higher resolution for the interval, including the Eocene-Oligocene transition. The samples (each 20 g) were carefully ground to pass through an 80 mesh and extracted with CHCl 3 in a Soxhlet apparatus for 72 h. The extracts were concentrated and separated into aliphatic, aromatic and polar fractions using chromatography on silica gel and alumina. After natural air-drying and dilution using CHCl 3 , the alkanes were analyzed using gas chromatography mass spectrometry (GC-MS). A blank sample was checked using the same procedures to monitor for contamination.
The analysis of the organic matter was carried out in the Key Laboratory of Gas Geochemistry, Institute of Geology and Geophysics, Chinese Academy of Sciences. GC-MS was performed using a Hewlett-Packard (HP) 5973 mass spectrometer in electron ionization mode (EI 70 eV) interfaced directly with a HP-6890 gas chromatography equipped with a HP-5MS fused silica column (30 m × 0.25 mm i.d.; 0.25 μm film thickness). The oven temperature was from 80 to 300°C at 5°C/min and held at 280°C for 20 minutes. Helium was the carrier gas.
Samples from the Xiejia section were chosen for spore and pollen analysis at 2-3 m intervals. About 150-350 g of each sample were treated with 36% HCl and 73% HF, respectively. Furthermore, the spores and pollen were concentrated after the remaining material was passed through a 10 μm mesh in an ultrasonic bath. The prepared specimens were mounted with glycerol and examined at 400 × magnification (detailed results are shown in another article 1) ).

n-Alkanes
(i) Distributions and contents. The distribution of nalkanes ranges from C 16 to C 35 , maximizing at C 29 or C 31 . In most samples, short-chain compounds (< C 21 ) were low or nearly absent, probably implying limited contributions from microbes, algae and bacteria in the playa system. A marked odd/even carbon predominance is observed above C 22 throughout the Xiejia section from 50.2 to 28.2 Ma ( Figure  3). High proportions of long chain C 27 to C 31 members are typical of terrestrial higher plants, where they occur as the main components of plant waxes. The Carbon Preference Index, CPI {[(C 25 + C 27 + C 29 + C 31 + C 33 )/ (C 24 + C 26 + C 28 + C 30 + C 32 ) + (C 25 + C 27 + C 29 + C 31 + C 33 )/ (C 26 + C 28 + C 30 + C 32 + C 34 )]/2} ranges from 1.45 to 6.23 (Table 1). These characteristics are consistent for all the depositional sequences in the section and in accordance with a terrestrial higher plant origin for the n-alkanes [33,[44][45][46].
The total content of odd-numbered long chain n-alkanes (C 27 + C 29 + C 31 + C 33 ) shows a generally declining trend from 1.33 to 0.05 µg/g with some oscillations with increasing depth throughout the section, corresponding to 50.2 to 28.2 Ma ( Figure 2; Table 1). The values remained relatively high and oscillate from 0.4 to 1.33 µg/g from 50.2 to 37.5 Ma. They begin to decrease very quickly after 37.5 Ma, ranging from 0.05 to 0.5 µg/g, and then remain very low for the rest of the section.
The C max variation throughout the section shows that C 31 dominates the sediments from 50.2 to 37.5 Ma. All the samples should expect grass to be dominant during this phase; C 29 n-alkane dominates from 37.5 to 32.7 Ma, suggesting a predominance of trees in the Xining Basin; a C 31 dominance occurs again during 32.7 to 28.2 Ma, implying a grass dominance. The C max shifts at 37.5 and 32.7 Ma might be related to vegetation changes occurring in response to climate change.
(iii) C 27 /C 31 , (C 27 + C 29 )/(C 27 + C 29 + C 31 ) and C 31 /(C 27 + C 29 + C 31 ) ratio. Modern molecule organic geochemistry has demonstrated that the ratio of C 27 /C 31 of n-alkanes is mostly indicative of the relative proportion of wood plants to grass vegetation [35,37,49]. The variation in the ratio of C 27 /C 31 n-alkane in the lacustrine sediments has been demonstrated to be related to the dominant land types shifting between woody plants and grasses [27,33].
For further comparisons, we use Q grass/plant = C 31 / (C 27 +C 29 +C 31 ) and Q wood/plant = (C 27 +C 29 )/(C 27 +C 29 +C 31 ), to represent the relative n-alkane abundance of grass and woody plants in the samples respectively, because the Q grass/plant and Q wood/plant indices have been widely believed also to represent the relative abundance of grass and woody vegetation [39,50]. In contrast, the value of C 31 /(C 27 + C 29 + C 31 ) is relatively high from 50.2 to 37.5 Ma, showing a higher proportion of grass relative to woody plants [39]. With the change in climate and environment, the values of C 27 /C 31 and (C 27 + C 29 )/(C 27 + C 29 + C 31 ) show a marked increase, whereas C 31 /(C 27 + C 29 + C 31 ) decreased significantly since 37.5 Ma, implying that the vegetation changed from grass to wood dominance. Then, C 27 /C 31 and (C 27 + C 29 )/(C 27 + C 29 + C 31 ) ratios stay generally high between 37.5 and 32.7 Ma, suggesting a predominance of trees over the Xining Basin.
At 32.7 Ma, the C 27 /C 31 and (C 27 + C 29 )/(C 27 + C 29 + C 31 ) ratios began to decrease again and stayed relatively low from 32.7 to 28.2 Ma, with the C 31 /(C 27 + C 29 + C 31 ) ratio increasing gradually, showing gradual development of grassland after 32.7 Ma.
(iv) Average chain length (ACL). The ACL values [(27C 27 +29C 29 +31C 31 +33C 33 )/(C 27 +C 29 +C 31 +C 33 )] indicate the concentration-weighted mean chain length of the C 27 to C 33 n-alkanes, and might reflect local temperature variation [51,52], as well as different vegetation types [32,41,42]. Plants in warmer/wetter areas have higher values than those in colder/drier areas [42]. The values decrease from 30.4 to 29.2, with some oscillation, with increasing thickness in the section (Figure 2; Table 1). They vary from 29.6 to 30.4 and stay generally higher during the 50.2 to 37.5 Ma period.  from 29.2 to 29.9 between 37.5-28.2 Ma, suggesting that the climate gradually became cooler from 37.5 Ma, resulting in woody plant development.

Discussion
From the results above, it is clear that these different n-alkane indices show a very similar trend from 50.2 to 28.2 Ma in the Xiejia section. They suggest that vegetation responded to climate related changes, notably temperature and humidity, in the Xining Basin. A better understanding of vegetation and climate variability in the Xining Basin can be gained by combining n-alkane proxies and palynological records for the same time intervals. According to the n-alkane and palynological proxies, the paleovegetation and paleoclimatic evolution of the basin can be divided roughly into three major phases. Figure 2 and Table 1 illustrate the variability in these records, showing three evolutionary phases of paleovegetation and paleoclimate, and two marked ecological shifts at about 37.5 and 32.7 Ma. Based on the dated Xiejia section, the three major phases are assigned to corresponding major marine isotope stages [53,54].
In Phase 1 (ca. 50.2 to 37. 5 Ma, samples XJ-1 to XJ-22), the distinctive predominance of C 31 and relatively high C 31 /(C 27 + C 29 + C 31 ) ratios might have resulted from a greater contribution from grasses, which tend to have longer chain compounds and produce a higher proportion of C 31 . We suspect that grasses or plants with waxy leaves that reduce water loss and have longer chain lengths were in greater abundance during the arid conditions of this phase. Thus, a higher proportion of C 31 during this phase is explained by aridity.
A warm/humid climate favors plant growth, which can be expected to result in a high proportion of odd n-alkanes.
In contrast, a dry and cold climate limits plant productivity and results in a low proportion of odd n-alkanes [38]. The odd n-alkane content is observed to be generally high and in fact reaches maximum values during this phase, suggesting relatively high plant productivity and a relatively warm or wet climate. Furthermore, the ACL values stay generally higher in this phase indicating a warm climate, which favors plant growth.
ACL, C 31 /(C 27 + C 29 + C 31 ) ratio and CPI show parallel changes throughout the section (Figure 2). The C 31 concentration significantly affects the value of CPI, showing a positive relationship. It seems that high CPI results mainly from the content of higher plant long chain n-alkanes. Hence, both CPI and ACL increase as C 31 /(C 27 + C 29 + C 31 ) increases. One explanation for these similarities is the relative increase in the C 31 contribution during warm stages, with consequent increases in ACL, C 31 /(C 27 + C 29 + C 31 ) and CPI. A similar trend was observed in the Chinese Loess Plateau profile [38]. Therefore, all three parameters indicate increased grass cover during warm stages.
The higher proportion of C 31 that occurred here most likely implies a warm climate. Other researchers have observed similar responses. For example, Hinrichs et al. [55] argued that the increase in C 31 n-alkane released from the western United States during MIS 5 could be interpreted as a response to increased temperature. Kawamura et al. [56] analyzed the n-alkanes from ten broad-leaf plant species collected from tropical (6°S), subtropical (27°N) and temperate (43°N) regions and found a systematic decrease in the relative percentage of C 31 from tropical to the temperate, also suggesting that C 31 is related to temperature. Thus, these n-alkane proxies are in accord with an obvious warm and dry climate during this period.
The pollen assemblage from this phase shows that Ephedripites and Nitrariadites were most abundant with some subtropical deciduous taxa (Figure 2), lending further support to the inference above a hot and dry climate in this phase. Generally, a hot-dry climate also favors grass growth. However, we do not observe high contents of grasses in our pollen spectra, except the inference of a high contribution of grass from n-alkane proxies above. Grasses identified from this part of the stratigraphy are largely Gramionidites, Sparganiaceaepollenites, Potamogetonacidites, and Cyperaceaepollis. These types of grasses have poor stratigraphic preservation. Pollens of grasses Chenopodipollis and Artemisia can be preserved in sediments for a long time, but they only appeared as important taxa in the ecologic system during the late Eocene and Oligocene [57][58][59]. We argue that these two points may have resulted in low grass contents in the early Cenozoic pollen record.
In Phase 2 (ca. 37.5 to 32.7 Ma, samples XJ-23 to XJ-37), the shift in the dominant peak from C 31 to C 29 might be related to vegetation changes that occurred in response to climate change. The C 27 /C 31 and (C 27 + C 29 )/(C 27 + C 29 + C 31 ) ratios largely increased, whereas C 31 /(C 27 + C 29 + C 31 ) and ACL values declined abruptly, indicating a predominance of trees over the Xining Basin. This suggests that the climate was relatively colder and humid during this phase. The long chain odd n-alkane component began to decrease very quickly from 37.5 Ma and the values remained very low for the rest of the section. The variation in long chain n-alkane content follows other n-alkane proxies, implying that 37.5 Ma may represent a marked ecological shift when the climate began to limit plant productivity, probably as a result of decreased temperatures.
This inference is supported further by pollen records. The most striking change is the sudden appearance and increasing abundance of cool-climate conifer pollens (in particular Picea) and grasses of Chenopodipollis and Artemisia during this period, along with the abrupt decline in warm-temperate and subtropical deciduous plants. After the sudden appearance of conifers, they constituted a major component of the palynological record. The relatively high abundances of Picea pollen occurred here, suggesting that the source plants grew in the vicinity of the study area at the time of deposition. Grasses of Chenopodipollis and Artemisia prefer a dry-cold climate and are demonstrated to appear mostly from the late Eocene to Oligocene [57][58][59], agreeing with our record. Thus, the pollen assemblage here indicates a vertical zone of vegetation occurrence during this interval, namely Ephedripites, Nitradites, Chenopodipollis and Artemisia growing at lower elevations, with warm-subtropical and warm temperate types at higher topographical regions, and cold temperate types only surviving at the highest elevations. It is estimated that forest or forest-steppe vegetation dominated the basin during this interval, with a spruce forest growing in the mountain areas, and drought-enduring shrubs and herbs spreading around the valleys, mountainous areas, glades and slopes. The climate was relatively humid and had a relatively low temperature.
A molecular fossil study of modern forest vegetation and underlying soil from the semi-arid area of Xinglong Mountain near Lanzhou in the western part of the Loess Plateau showed that branches and leaves of deciduous broad-leaf trees are dominated by C 27 homologues whereas needle-leaved evergreen trees and leaves are dominate by C 29 homologs [39]. The greater abundances of these C 29 n-alkanes compounds from the Xiejia section from 37.5 to 32.7 Ma may reflect a greater presence of needle-leaved evergreen trees (especially Picea). The n-alkane proxies agree well with the explanation of the pollen diagrams, suggesting a predominance of trees and a cold and wet climate during this phase.
In Phase 3 (ca. 32.7 to 28.2 Ma, samples XJ-38 to XJ-43), C 31 again dominates. C 31 /(C 27 + C 29 + C 31 ) largely increases, whereas C 27 /C 31 and (C 27 + C 29 )/(C 27 + C 29 + C 31 ) ratios decrease sharply and ACL values stay relatively low after 32.7 Ma, indicating a predominance of herbaceous plants and a relatively cold climate, respectively. The odd n-alkane component is generally low and, in fact, drops to its lowest value (~0.05 µg/g). The lowest content, which occurred after ca. 33 Ma around the Eocene-Oligocene climate transition, implies that the cold climate limited plant productivity. Thus all these n-alkane proxies suggest the climate was probably much drier and colder than before.
The pollen record shows that drought-enduring shrub and herbaceous plants, especially Chenopodipollis and Artemisia, grew luxuriantly, whereas coniferous, temperate and warmtemperate deciduous components declined abruptly after ca. 33 Ma. Steppe vegetation replaced mountain spruce forests, suggesting the development of an arid climate and intensifying aridification.
Sediment lithology and facies provide further support for this inference. Alternate layers of gypsum and gypseous mudstone of playa were replaced by red beds of siltstone and mudstone from the flood plain at ca. 33 Ma (Figure 2), suggesting a complete drying-up of the playa by a drier climate.
Thus, organic geochemistry and pollen-spore records collectively demonstrate a long term cooling trend from 50.2 to 28.2 Ma and an ecological event at 37.5-32.7 Ma when a vertical zonation of vegetation from lowland arid grasses, to subtropical deciduous plants at moderate elevations, to high-elevation coniferous trees. We think these changes in climate and vegetation were responses to the long term global cooling since the Eocene climatic optimum at ca. 52 Ma [12,54] (Figure 2) and probably the uplift of the surrounding mountains on the northern part of the Tibetan Plateau in the early Cenozoic. Recent tectonic studies and basin analyses do show that the Altin Tagh, Qilian Mt. and east Kunlun Mts. had an early episode of rapid uplift around 40-30 Ma [60,61]. Whereas the dry event at ca. 33 Ma was probably a direct response to the Eocene-Oligocene climate transition due to the initiation of Antarctic glaciation [12,54].

Conclusions
(1) The plant n-alkane proxy records obtained from early Cenozoic playa and fluvial stratigraphy spanning 50.2-28.23 Ma in the Xining Basin show three evolutionary phases of paleovegetation and paleoclimate with discrete changes occurring at ca. 37.5 and 32.7 Ma. The first (50.2-37.5 Ma) and second (37.5-32.7 Ma) phases are characterized by high and low values of C 31 /C 27 , ACL, CPI and n-alkane content respectively, whereas the third phase (32.7-28.23 Ma) shows a similar ratio of C 31 /C 27 to the first phase, but is accompanied by low values of ACL, CPI and n-alkane content. These changes clearly demonstrate three phases of climatic and ecologic variations.
(2) The palynological records also demonstrate three phases of changes similar in age to the n-alkane proxy records presented above. The first phase presents a predominance of Ephedripites and Nitrariadites with some subtropical broad-leaf trees belonging to many taxa. The second phase is characterized by the sudden and spectacular appearance of conifers and shrub-grasses (Artemisia and Chenopodipollis) and a large reduction in subtropical broad-leaf trees. The third phase presents a predominance of Artemisia and Chenopodipollis and clear reductions of Ephedripites and Nitrariadites and conifer trees. These changes in vegetation indicate not only three phases of climatic change, but also that the vegetation had developed a distinct vertical zonation with arid grasses in the lowlands, subtropical deciduous plants at middle elevations, and coniferous trees in the high mountains from the second phase.
(3) Our n-alkane and palynological records collectively demonstrate that the paleoclimate in the Xining Basin experienced a long term cooling trend from 50.2 to 28.2 Ma with a distinctive ecological event spanning 37.5-32.7 Ma when a vertical zonation of vegetation was established. The long term global cooling since the Eocene climatic optimum and uplift of the surrounding mountains in the northern part of the Tibetan Plateau in the early Cenozoic may be responsible for these changes.